Joint Chromatic-Aberration-Correction And Super-Resolution For Computational Spectral Imaging

Spectral imaging has been widely used in meteorological observation,land planning,medical diagnosis,military investigation,disaster prevention,etc,due to its capability of characterizing the components and structures of the scene.Recently,inspired by the compressive sensing theory(CS),Duke University proposed a new computational spectral imager called coded aperture snapshot spectral imagers(CASSI),which can capture the power spectral density of light via single shot.In general,the whole spectral imaging process is divided into two stage: the observation process and the data restoration process.In the observation stage,the measurements are obtained by coding and sampling of the spectral scene.In the data restoration stage,the 3D spectral data cube can be restored by solving a CS inverse problem.However,due to the limitation of material and craftsmanship,There are some drawbacks in the observation process,such as the loss of spectral information,the low sampling rate and chromatic aberration,may bring serious problems to CASSI system,like poor imaging quality and low resolution.Based on this,some papers put forward specific solutions for better reconstruction quality,such as multi-frame observation and an optimization algorithm based on adaptive dictionary,although the above strategies have achieved good results,there are still some space for improvement.The effect of chromatic aberration in CASSI system leads to varying degrees of nonlinear distortion in the process of sampling,which seriously affects the reconstruction quality.In this paper,we investigate the chromatic aberration in CASSI system,especially for problem of how to the reconstruct the spectral image via the chromatic aberration measurements.Firstly,we model the chromatic aberration as a point-spread-function(PSF)processing after coded aperture,and embed the model into the CS reconstruction process.Considering the highly nonlinear distortion of the CASSI system,the PSF varies significantly all over the image and it is intractable to estimate the PSF one pixel by one pixel.To tackle this dilemma,we propose a piecewise linear approximation strategy for the PSF estimation.The second contribution of this paper is to improve the resolution of traditional CASSI system.We involves another panchromatic channel into the CASSI system.In addition,the low resolution coding aperture of CASSI system is replaced by a high resolution version.Then the reconstruction problem is to convert to restore the high resolution spectral image via high resolution coding aperture and low resolution sensors with the side information which is provided by the panchromatic image.Experimental results shows that the proposed strategy can significantly improve the quality of CASSI system in the sense of clear edges,fine textures and higher resolution.